Bayesian apologetics: 'Estimating the Evidence'
Interpreter's 23-part Bayesian series shows the form of rigorous reasoning, but its conclusion is fixed by subjective likelihoods and by which evidence is counted as independent.
What the Church claims
A fair probabilistic weighing of all the evidence strongly favors the Book of Mormon's authenticity.
Once the evidence is in, this claim: Weak. The methodological critique is sound: the series treats correspondences as independent when they are not, and the priors end up doing the work the evidence is supposed to do. The final numbers cannot be checked while the pages are offline and no formal rebuttal has been published, which is why this is weak rather than untenable.
The record
- Kyler Rasmussen's 'Estimating the Evidence' ran on the Interpreter Foundation site in 2021 as Episode 0 ('On Quantifying Skepticism') through Episode 23 ('On Conclusions'), each assigning likelihood ratios to a line of evidence (e.g. dictation speed, Early Modern English, geography, 'Late War' parallels).
- Reader comments on the Interpreter pages contested specific numbers, e.g. that Early Modern English 'is as likely to be evidence of modern fiction', and one respondent disputed an episode that scored five orders of magnitude against the book. These come from excerpts of the comment threads; the episode pages themselves no longer load.
The moving target
The Church has never endorsed a Bayesian case for its truth claims, and has never published a probability for any of them. Its own standing instruction runs the other way: the Book of Mormon asks the reader to pray about it, not to compute it. The series below is the work of one independent writer, and nothing the Church teaches depends on its arithmetic.
The exchange
Kyler Rasmussen, Interpreter Foundation (2021)
Start with an extremely skeptical prior against the Book of Mormon, then update on each independent line of evidence with conservative likelihood ratios; the cumulative posterior still ends up favoring authenticity.
The answer: Bayes' theorem is only as good as its inputs. (1) The likelihoods - 'how probable is this feature if Joseph Smith or a nineteenth-century author produced the book?' - are the analyst's estimates, and small changes compound across 23 multiplications into dozens of orders of magnitude. (2) Multiplying ratios assumes the lines of evidence are independent; many (chiasmus, Hebraisms, EModE, complexity) are correlated features of a single biblically saturated dictation. (3) The reference class matters: asking how likely it is that an uneducated farmer would produce X ignores the base rate of unexpectedly capable religious texts (the Quran, Oahspe, A Course in Miracles, The Urantia Book). (4) Strong negative evidence (DNA, anachronisms, Deutero-Isaiah, KJV errors) has to be discounted with auxiliary hypotheses whose own prior cost is not always charged. The series is best read as a transparent statement of one believer's credences, not as a measurement.
What they get right: Rasmussen deserves credit for stating numbers that can be challenged, for including evidence against, and for beginning with a skeptical prior - more than most critics or apologists do. Critics should engage the inputs rather than mock the exercise.
Kyler Rasmussen, Interpreter Foundation
Rasmussen published his priors, his likelihood ratios and his reasoning across all 23 episodes. Anyone who thinks the numbers are wrong can substitute their own and show the result.
The answer: Publishing the inputs is better than hiding them, and it still does not make the exercise informative. A likelihood ratio means something only when it is anchored to something outside the analyst — a base rate, a reference class, a second rater working blind — and none of the 23 episodes supplies one. The invitation to publish your own numbers concedes the point: if the answer swings from overwhelming support to overwhelming refutation depending on who assigns the ratios, the calculation is measuring the analyst rather than the evidence. It borrows the authority of measurement without doing any measuring.
What they get right: Making assumptions explicit is better practice than the impressionistic 'weight of evidence' appeals it replaced, because an explicit assumption can be examined.
Both cases at full strength
The skeptics' case
The same method with different but equally defensible likelihoods yields overwhelming odds against. That is the point: when the output swings by dozens of orders of magnitude depending on subjective inputs, the formalism is not doing the epistemic work. Better practice is to publish ratings with explicit sensitivity ranges and to show, for each item, what both sides would have predicted in advance.
The apologetic case
All historical reasoning is implicitly Bayesian; making it explicit exposes assumptions to criticism. Rasmussen set a deliberately hostile prior and still found the evidence for authenticity cumulatively heavy; critics who dislike the numbers should publish their own.
What would change this
An adversarial collaboration in which a critic and an apologist agree likelihood ranges in advance.
Closely related
Sources
- Estimating the Evidence (Kyler Rasmussen's 23-part series)
- Estimating the Evidence, Episode 0: On Quantifying Skepticism
- Estimating the Evidence, Episode 23: On Conclusions
An independent project. Not affiliated with The Church of Jesus Christ of Latter-day Saints or with any organization cited here. Summaries are written here; short quotations are attributed and linked. Every page names its sources so that it can be corrected, and every page can be challenged.